DRAM prices expected to double in Q1 as AI ambitions push memory fabs to their limit
Summary
Industry analyst TrendForce has revised its forecasts sharply higher: DRAM contract prices are now expected to jump about 90–95% quarter‑on‑quarter in Q1 2026 (nearly double), while NAND flash is forecast to rise 55–60% in the same period. The revision reflects surging procurement from AI‑focused hyperscalers and cloud service providers, plus stronger‑than‑expected PC shipments in late 2025.
TrendForce warns LPDDR (used in notebooks, phones and increasingly in high‑end AI racks) will see roughly 90% QoQ price rises — some of the steepest hikes on record. Hyperscalers are also buying enterprise SSDs in volume to support inference workloads, further tightening NAND supply.
The immediate causes are demand for memory to store LLM inference key‑value (KV) caches and a broader shift from training‑heavy buying to inference provisioning, which requires more DRAM and storage. Although vendors are investing in new fabs, capacity additions will take years, so prices are expected to remain elevated through 2028.
Key Points
- TrendForce now predicts DRAM contract prices will rise ~90–95% QoQ in Q1 2026 — effectively doubling from Q4 2025 levels.
- NAND flash is forecast to increase by about 55–60% QoQ as CSPs and hyperscalers rush to buy SSDs for inference workloads.
- LPDDR4x and LPDDR5x pricing is expected to climb roughly 90% QoQ — the steepest increases in their history.
- AI inference growth (storing/precomputing KV caches) is driving much of the extra memory and storage demand.
- OEMs typically buy memory in bulk ahead of demand; as inventories run down and restocking happens, system prices (not just module prices) are likely to rise.
- New fabs and capacity will take years to come online, so relief is not imminent — prices forecast to stay high through 2028.
Why should I read this?
Short version: if you buy servers, configure clouds, buy laptops or manage infra budgets, this is going to hurt your wallet. Memory prices are surging because AI players need buckets of RAM and SSDs — so expect higher kit costs, squeezed procurement windows and painful budgeting choices. We’ve skimmed the numbers and pulled out the bits that matter so you can decide whether to buy now, delay or plan for sticker shock.
Author’s take
Punchy and plain: this is more than a blip. The memory squeeze is structural and AI is the accelerant. If your capex calendar is flexible, consider accelerating critical purchases; if not, brace for higher TCO across servers, laptops and edge devices.